AI artist Yasmin Gross also known as x.machina.flora is German-born and Paris-based, with a graphic design background and more than ten years in luxury visual communication. Her AI art moves between fashion, flowers and fine jewelry. The work is varied, yet a sure and assertive sense of taste runs through every series.

We spoke with her about her path, the artists who feed her eye, and what she would tell someone starting out in AI image generation.

Here is our conversation, in her own words.

Could you introduce yourself?

I’m Yasmin Gross, an AI artist and art director based in Paris. I’m German, from Frankfurt, and I’ve lived abroad for over fifteen years, first in the UK and then in France. I trained as a graphic designer, then spent over ten years in luxury visual communication, working across photography, set design and art direction, before AI became part of my practice.

Today I work under two names: my own, Yasmin Gross, and X Machina Flora, my personal AI generative practice, where floral structures merge with distorted human forms. More recently I co-founded VTOPYA with Vincent Verdureau, a creative studio producing AI visuals for fashion, beauty and fragrance brands. Everything feeds into everything else: the personal work is where I experiment freely, and the studio is where those discoveries meet real briefs and real constraints.

How long have you been using generative AI?

Since around 2023, and it became professional very quickly. I was almost addicted to feeding my curiosity and generating my ideas with AI, and X Machina Flora became the outlet. Through it I got to work on many different projects, solo or in larger teams, from beauty to the music industry. In 2025 Vincent and I founded VTOPYA, and now AI is part of my daily work for brands, agencies and smaller creatives.

Which AI tools do you use the most?

Midjourney is still my main tool for stills, it’s where the image is born. Then Nano Banana Pro for finishing and product integration, Magnific and Topaz for upscaling, Kling and Luma when an image needs to move. And Photoshop, always. Nothing leaves the studio without passing through it.

I write prompts like a photo briefing: subject, framing, lighting, material, palette, era or reference, then render settings. Sometimes I add poetic words or concepts that don’t fit the classic studio setting. They give the machine room to guess, and it often guesses wrong, but when I find the right combination I know I’m onto something.

A single word can derail everything. “Dreamy” blows out the light, and once “winged” gave a perfume cap literal angel wings. When something keeps going wrong, I name the exact defect I’m seeing in the prompt instead of writing generic negatives. It works much better.

I keep style reference weights very low. It’s a bit of a pile-up, so I always start at the bottom and build from there. A strong style reference flattens everything into the same image.

I’m currently training Midjourney moodboards on my own archive, my AI work and my photography, to build styles that come from me rather than from the model’s defaults. I never feed them with other people’s images. I only use my own, or images a client has given me for their project, and otherwise I work with prompts alone. We already don’t know what these models were trained on, so I want to keep my own process as morally correct as possible. At VTOPYA we’re also training our own models, and I hope that eventually we can become platform independent.

And when AI simply can’t do it, I stop fighting. I generate clean elements separately and composite them by hand. Knowing when to leave the machine is part of the method. If it doesn’t work one way, there’s always another.

Where do you draw your inspiration from?

Nature first, flowers especially, and the body. I’m fascinated by the moment where something organic stops being one thing and becomes another, a petal that turns into skin, a face that opens like a bloom. My years in beauty and fine jewelry imagery also left a mark: I love precise light, texture and material, things that feel almost touchable.

But really, I take inspiration from everywhere. I love illustration and scientific drawings, but also anime, fashion, cinema, contemporary art, photography, fables and stories. I sometimes feel like a sponge, I can’t help being influenced by everything I’ve seen and loved, and I don’t discriminate in taste. I’d describe my personal style as retro-futuristic and always maximalist.

How do you organize your visual research?

My own archive is my most important reference. After a few years of generating, you accumulate thousands of images and prompts, and going back through them is often more surprising than looking at someone else’s work. I keep them indexed and searchable so I can find a texture or a light I made two years ago. I also get inspired by fellow artists and current trends, which I then transform and adapt to my own universe.

Are there any non-AI artists who influence your work?

So many. In fashion, Nick Knight for the way he pushed technology into image-making long before AI, Elsa Schiaparelli for her surrealism, and Alexander McQueen and Iris van Herpen for how they bring nature onto the body. Tim Walker too, for turning fashion into fairy tales.

Anime shaped my eye early: Akira, Perfect Blue, Ghost in the Shell, bodies that transform and identities that blur. I love Hajime Sorayama’s chrome retro-futurism and the costumes of Eiko Ishioka.

And then the botanical side: Ernst Haeckel’s drawings, Karl Blossfeldt’s plant photographs, and Maria Sibylla Merian, who happens to come from Frankfurt like me. I’m also drawn to other cultures and their visual traditions, Japanese prints, Persian miniatures, textiles, anything with layers of pattern and story.

Do you have a story to share about an idea that came from an unexpected source?

One of my favourite projects came from a dialogue with the Tokyo calligrapher Ken Wakabayashi. I created AI images inspired by calligraphy, he answered them with his own handwritten pieces, and I scanned his work, re-textured it with AI and layered everything back together in Photoshop. It became a conversation between his hand and the machine, across two cities.

Ken Wakabayashi‘s website
Ken Wakabayashi’s INSTAGRAM

What advice would you give to someone starting out today?

Learn to see before you learn to prompt. Creating with AI involves a lot of editing yourself. When everyone can create “almost” the same thing, what gives your creation a purpose? Is it practice for a future client? Were you simply inspired? I find that figuring out the purpose behind your creations adds variables and frameworks to your work.

The people who get the most out of these tools are the ones who follow their own intuition. I’ve always loved languages and found it relatively easy to pick up vocabulary, which is probably one of my strengths in this industry. Practice describing things, and see if someone else understands your vision without seeing it. That usually shows you pretty quickly how clear your idea really is.

I sometimes rework ideas a year later, and it’s exciting to see how much either the AI model or my ideas have evolved.

And don’t chase every new tool. A new model comes out every week. Don’t chase the hype either, because chances are the idea has already been done. If it has, find a new way of showing it rather than recreating it.

That also comes with a responsibility. Not copying someone else’s work has always been a basic rule in every creative field, and it shouldn’t change just because AI makes it so easy. Being inspired by someone is one thing, reproducing their work is another, and that line is still ours to hold.

If you had to start over, what’s the first thing you would do differently?

I would document everything from day one. Every prompt that worked, every word that broke an image. It took me a while to understand that this knowledge is the real asset, much more than any single image.

Any mistakes you’d be willing to share?

Letting AI write too many of my concepts. When nothing in the prompt is written by you, it’s hard to figure out where an error comes from, and it takes even longer to understand the balance of your images: how much comes from the prompt, how much from the references, how much from the settings. Now I prefer to build base prompts and reuse them, changing only the setting or the characters, rather than starting from scratch every time.

Another one was testing too many platforms, and chasing other people’s quality and results. You never really know how someone else created an image, what they edited, composited or retouched afterwards, so comparing yourself to the final result can send you in the wrong direction.

Any unexpected tips about tools, methods, or your daily practice?

Spend as much time looking as generating. Most of my progress came from sitting with a batch of images and asking precisely why one works and the others don’t.

Any advice on building a social media presence?

Build a universe, not a feed. People follow a recognizable world, a sensibility they can come back to. Having X Machina Flora as a separate identity helped me a lot: it gave the work its own space to exist and grow. Be consistent, and post what you love rather than what you think will perform.

A piece of advice you wish you’d received earlier?

That your taste is the only thing a machine can’t give you, so protect it and train it.

And stop thinking that creating good AI work is fast. You might need fewer people, but from my experience in post-production and in the industry in general, the shoot was never all that’s behind a photo. With AI it’s the same: the generation is only the beginning, and most of the work happens after.

Images © x.machina.flora 2026